Justify, Design, Mitigate, Inform – is that how we should use facial recognition?
Facial recognition products created by leading technology vendors such as Amazon, Google, Dell, and Microsoft are easy to implement.
In recent times, the use of the technology by law enforcement and global conglomerates has been criticized on the grounds that it erodes public trust and infringes on citizens’ and employees’ rights to privacy.
That being said, facial recognition has also delivered results that cannot be ignored, especially to security professionals – law enforcement or otherwise – looking to manage large crowds.
Further, in the current environment with the coronavirus outbreak, facial recognition is being used by China and others to prevent the spread of the disease and protect the wellbeing of people everywhere.
Clearly, despite the uproar against facial recognition, mostly from uninformed citizens, there are benefits to augmenting and scaling up its use.
To make the transition into a facial recognition-powered world responsibly, the World Economic Forum (WEF) has put together a set of four best practices to keep in mind when designing such solution.
# 1 | Justify the choice of facial recognition technology
“This implies defining the problem to be solved and explaining how facial recognition technology might better solve this specific problem compared with alternative technologies,” explains the WEF’s document.
Given the ease of deployment and the low cost of implementation, the use of facial recognition technology is growing fast. However, in many cases, its use could be excessive. That’s what the WEF wants to draw attention to and avoid in coming months as spending on facial recognition grows.
# 2 | Design a data plan that matches with end user characteristics
“Based on the defined characteristics of the end users, a data plan needs to be designed that includes fairly equal samples of these subgroups and collects data accordingly.”
Since facial recognition offerings are becoming common, the WEF reminds those looking to acquire and implement the solution that testing it using appropriate data in the right conditions is key.
Imagine using a solution that was developed to recognize members of a club and provide them with access to facilities while they were in a well-lit premise such as a clubhouse. That same solution will pose challenges if rolled out in a dimly lit parking lot and hence, needs to be trained using better data.
# 3 | Mitigate the risks of biases
“Define the risks of unfair biases in the system to be developed for flow management use cases.”
According to the WEF, it is important that businesses rolling out facial recognition evaluate each step in their process of use, document the characteristics of end users to identify and eliminate the risks of discrimination, and finally define the environment in which each of the identified risks will be evaluated.
When businesses pay attention to these factors and work on mitigating them, they’ll naturally be more mindful of how solutions work to best serve users.
# 4 | Inform end users, and be transparent
The WEF understands that the biggest issue with facial recognition is the lack of transparency.
Hence, the guidelines suggest that governments and businesses using the technology should not only ensure that everyone has access to information about the functioning of the system, but also the principles that guided its design and use.
Further, the WEF recommends adopting a consent policy that includes intended use, data retention periods, data protection and sharing policies, and so on — all of which gives people more control over the use of the technology and over time, helps get their support.
Although the WEF clarifies that the best practices suggested apply more closely to ‘flow management’ use cases, as against facial recognition-based payments or healthcare solutions, there’s nothing stopping more developers from adopting these as a ‘bare minimum’ criteria going forward.
Governments deploying the technology to protect and serve its citizens might determine the criteria set by the WEF on a case by case basis but businesses could definitely take a steer from the WEF’s latest document when rolling out facial recognition-powered solutions for internal and external use.